基于频谱的茶叶种类识别方法研究

Research on the Identification of Tea Based on Frequency Spectrum

为准确、快速、无损的识别茶叶种类,以3种不同的茶叶图像为研究对象,首先,通过颜色统计分析、阈值分割和形态学处理获取茶叶目标图像;其次,对茶叶灰度图像进行二维傅里叶变换后,将图像功率谱分成20个等间距同心长方环,提取每个长方环内功率谱能量占总能量的比值作为茶叶特征参数;最后,利用交叉验证结合支持向量机实现了茶叶种类的自动识别,平均识别率可达86.7%。

To accurately,rapidly and nondestructively identify the species of tea,a method for identification of tea species was proposed by using image processing and image spectra analysis.Firstly,the tea images were obtained through statistical color analysis,threshold segmentation and morphological operations.Secondly,two-dimensional Fourier spectra for each tea image were extracted through Fourier transform,and then the image power spectrum was divided into 20 equidistant concentric rectangle loops.The ratio of the power spectrum energy within each rectangle loop compared with the total energy was calculated,and the obtained value was used as a parameter of tea features.Lastly,the species of tea was realized automatically by cross validation and Support Vector Machines.The experiment showed that the average recognition rate of tea images was 86.7%.